Hacker News Reader

Comment by weiyin | Hacker News Reader

Hi, thanks for checking us out. We're happy to answer any questions in the comments.

Some of you may be familiar with app.io's demo pages service which was discontinued in August. We loved using their service and we thought it'd be a blast building a replacement.

Technologies used: node, socket.io, mongodb, html5 canvas

View on HN

I understand those are the web technologies, but how is this being run in the back? Tons of mac servers, with multiple IOS simulators on each? One running per user on the mac? Custom software to control and pass through to the socket?

Care to shed any light as to an overview on how the backend works? I've always been curious since I first saw App.io

> Tons of mac servers, with multiple IOS simulators on each?

Maybe a modified iOS Simulator binary that can run hundreds of instances on a single powerful server, would be my best bet.

Maybe, but then launching multiples and capturing the screen of them to send off through the sockets seems like there must be something in between. I wonder if its able to see where the session windows are on the screen and screenshotting them to send through, or if they somehow hooked directly in to pull the screen?

Hey B-n-c, You are pretty much spot on with everything you mentioned.

Apparently app.io had some setup w/ VMs running on lightweight hypervisors (https://macstadium.com/casestudies), but we found we didn't have to have all that overhead to run multiple simulators per machine.

And you're exactly right about the small piece of custom code to pass controls and send frames. We're big fans of socket.io - pretty powerful stuff!

Happy to answer follow-ups if we missed something.

Thanks, weiyin

The one piece that puzzles me is how the frames are being captured from the iOS simulator. Are you launching multiples per user account, where each one is a different session at a different portion of the screen, and essentially screenshotting each? Or did you modify the simulator somehow and are hooking in to grab the screen, so that way its irrelevant of their position on screen?

Thanks!

We wrote some custom OS X code to capture frames from individual processes. That way the simulators don't need to be positioned in any way or even visible. Cheers.

The other piece of magic you are doing is sending in the keystrokes / mouse movements. If you're running multiple iOS simulators on a Mac, how are the key presses / mouse movements not interfering with other users?

(don't see a link to reply under your last comment)

yea, so the incoming keystrokes and mouse movements are managed through some custom OS X code as well.

the specific method we're using is CGEventPostToPSN, more information here: https://developer.apple.com/library/mac/documentation/Carbon...

(No reply button under your other comment either)

Thats a pretty sweet API, never seen that before, thanks! I'm going to play around with it a bit to see what it can do!

As previously the Head of BD at App.io, I applaud your efforts! Well done! Quite a few ways you can monetise, good luck! :)

Likewise!! Well done guys. Great to see you've grabbed the torch. Good luck and I look forward to following your progress.

Thanks, it means a lot to us coming from you guys!

Can't wait until you figure out h.264 and the MediaSource API ;) You guys are moving at an incredible pace! Very impressed.

ps. I love the new /embed endpoint and iPad support. Keep it coming!

Reply on news.ycombinator.com